Entity search: How to build virtual documents leveraging on graph embeddings

We develop an innovative method for the Information Retrieval Entity Search task. We propose a new approach that exploits graph embedding techniques and clustering in order to create the documents necessary for the retrieval, in particular we create a document for a set of related entities. The main advantage of our implementation is that our systems could return to the user not only entities that directly match the user query, but also relevant entities that are not explicitly mentioned.